Work Package 4: Distant-speech and speaker recognition
This work package regards the development of robust solutions for speech recognition and speaker identification based on distant-talking interaction. Beside a progress in each of the reported topics, a main target is to identify the ASR solutions to adopt for real-time prototypes in the given languages.
The following tasks are present in this work-package:
- T4.1 – Acoustic and language modeling – baseline development for different languages and its evolution
- T4.2 – Acoustic features
- T4.3 – Adaptation to speaker and environment
- T4.4 – Confidence measures, keyword spotting and rejection strategies
- T4.5 – Distant speaker verification and identification